Abstract
In this thesis, we use textural features to retrieve images from a database. Here, the features we used include Fourier, Gabor, and Wavelet features. A fast search method was proposed. It not only saves time but also improves the correct retrieval rate. The Whitney feature selection method is adopted to select good features and Learning Vector Quantization (LVQ) clustering method is used to cluster all the images of the databases.Two databases were established in this thesis. The first database is created with 960 128’128 texture images scanned from Brodatz’s texture book. It includes 60 classes, and each class contains 16 images. The second database consists of 150 256’256 images. It includes face, fingerprint, city, and two kinds of seal photos, 30 images for each category. The experiments are done on both databases to demonstrate the effects of the proposed method.